Fewer

Abstract

In federated learning, the local devices train the model with their local data, independently; and the server gathers the locally trained model to aggregate them into a shared global model. Therefore, federated learning is an approach to decouple the model training from directly assessing the local data. However, the requirement of periodic communications on model parameters results in a primary bottleneck for the efficiency of federated learning. This work proposes a novel federated learning algorithm, Federated Weight Recovery(FEWER), which enables a sparsely pruned model in the training phase. FEWER starts with the initial model training with an extremely sparse state, and FEWER gradually grows the model capacity until the model reaches a dense model at the end of the training. The level of sparsity becomes the leverage to either increasing the accuracy or decreasing the communication cost, and this sparsification can be beneficial to practitioners. Our experimental results show that FEWER achieves higher test accuracies with less communication costs for most of the test cases.

Characteristic Changes in First-Visit Patients with Peripheral Facial Palsy before and during COVID-19 Pandemic: A Single Center Retrospective Study

“…Patients in their 40s-50s accounted for the highest proportion in 2019 and 2020; however, those over 60 accounted for the highest proportion in 2021. These results are consistent with the findings of a previous study that showed that the medical utilization rate among those under the age of 18 and those in their 20s-40s decreased significantly during the COVID-19 pandemic, while the rate of medical utilization rate among those aged 60 and older slightly decreased [32]. This result can be attributed to several complex factors, such as the tendency of older adults to maintain medical care utilization despite the pandemic, underestimation of subjective health conditions, and increase in the number of the population aged over 65 years every year [22,33].…”

VCAb: A web-tool for structure-guided antibody engineering

“…We note that for Rosetta, since each point mutation was predicted separately, the resultant scores on VCAb were not additive and therefore would not be suitable for predicting the effect of multiple mutations in combination. Here, AntiBERTY, being a language-model based method, can address this issue: since each amino acid is represented as a “word” in a sentence, the pseudo-log-likelihood returned by the model can be summed together to represent the likelihood of observing several amino acids in combination (Figure S4, Supplementary Materials), in the same way as the likelihood of a given sentence being presented is evaluated in language models used in natural language processing (NLP) (Meier et al ., 2021; Salazar et al ., 2020; Shin et al ., 2019). mAb114 is an antibody binding to the glycoprotein of ebolavirus (PDB ID 5fha).…”